Embodiments of this invention relate generally to providing care to patients in healthcare locations. More particularly, embodiments of this invention provide a system and method for creating treatment plans for patients that are continuously updated based on patient data.
Advances in communications, video displays, monitoring devices and computers have made it possible to remotely monitor hundreds of monitored patients. Alerting systems may be deployed to alert healthcare providers when certain conditions are met. For example, in U.S. Pat. No. 5,942,986 issued to Shabot, et. al for a “System And Method For Automatic Critical Event Notification,” describes a critical event notification system that permits review of a patient's diagnostic information, lab results, chart, or other data, automatically, by computer or similar equipment, and it provides for automatic paging of a responsible physician or physicians should a “critical event” be detected. The decision to page an individual is made automatically by the system, and does not require a direct human decision.
“Decision Support Systems in Critical Care” (Edited by M. Michael Shabot and Reed M. Gardner, 1994), is a compilation of articles that collectively describe the application of computers in health care settings. Decision support systems are defined as systems that receive medical data as input and produce medical information and/or knowledge as output. In some implementations, decision support systems utilize inferencing methods to detect associations between different pieces of information, alerting clinicians to certain patterns of events, which may be serious or life-threatening.
An example implementation of an inferencing method is described in the context of analyzing blood gas readings and laboratory results. Three different types of alerting algorithms are described: 1) high and low critical values 2) calculation-adjusted critical values, and 3) critical trends. (See, Decision Support Systems in Critical Care, pages 157-65.) The calculation-adjusted critical value algorithm reflects the dependence of the algorithm on multiple parameters. The application of the inferencing module produces an alert that is displayed on a screen or sent to a wireless device.
In U.S. Pat. No. 6,804,656 issued to Applicants, a smart alarm system was described. The smart alarm system of the '656 Patent, constantly monitors physiologic data and all other clinical information stored in the database (labs, medications, etc). The rules engine searches for patterns of data indicative of clinical deterioration. By way of illustration, one family of alarms looks for changes in vital signs over time, using pre-configured thresholds. These thresholds (also referred to as “rules”) are patient-specific and setting/disease-specific. Physiologic alarms can be based on multiple variables. For example, one alarm looks for a simultaneous increase in heart rate of 25% and a decrease in blood pressure of 20%, occurring over a time interval of 2 hours. Alarms also track additional clinical data in the patient database. Other rules follow laboratory data (e.g. looking for need to exclude active bleeding and possibly to administer blood). Regardless of the data elements that are used, the purpose of the rules is to facilitate detection of changes in a patient's condition (whether that condition is improving or degrading) in a predictive manner and to automate a response appropriate to the “new” condition.
Clinical prediction has been practiced in various forms since the first doctor practiced medicine. Observation, intuition, and the prevailing wisdom of the time were used to diagnose and treat illness. The patient's status following a particular treatment was associated with the treatment, whether or not there was a causal connection between the two. Often, treatments that were ineffective, or worse, deleterious, were perpetuated because there was not basis for determining whether there was a cause and effect relationship between the treatment and the result.
Scoring systems were developed to identify the important physiologic parameters and chronic health conditions that determine clinical outcome. The typical clinical prediction rule is geared to determine a specific outcome. The identification of the key predictive variables is accomplished using well-known statistical techniques. The model is validated by applying the scoring system to patients and confirming the outcome against the predicted outcome.
An example of a predictive scoring system is the Acute Physiologic and Chronic Health Evaluation II (APACHE II) instrument commonly used to assess patients for admission to an ICU. The APACHE II instrument studied 5800 patients, and 13 hospitals, and, with statistical methods, identified 12 continuous physiologic variables measured within twenty-four hours of ICU admission.” These variables were coupled with others describing the chronic health of the patient. APACHE II has also been applied to a wide variety of clinical issues in critical care and has been the method of choice for describing the severity of illness in some landmark studies.
APACHE II has several unintended flaws. The first was that the derivation data set was relatively small and therefore did not have the statistical power to describe subsets of disease, such as congestive heart failure, liver failure, and hematologic malignancy. The second was that the instrument could not distinguish between patients who had prior treatment and those who did not. This flaw, now termed “lead-term bias,” was discovered when a number of investigators demonstrated that the predictive accuracy of APACHE II faltered when it was applied to patients who were transferred from other ICUs or from within the hospital setting. In these situations, the APACHE II instrument underestimated mortality.
APACHE III was intended to correct this deficiency. This instrument was derived from 17,440 patients, and 40 hospitals, representing a wider spectrum than APACHE II. APACHE III employs 17 continuous physiologic variables, chronic health information, prior treatment location before ICU admission, and principal ICU diagnosis. It also has a new feature whereby mortality prediction is updated on a continuous basis.
A Simplified Acute Physiologic Score (SAPS) was developed to streamline the approach utilized by the APACHE systems. The SAPS II system employs 17 variables: 12 categorical physiologic variables; age; type of admission; and three other designated disease variables (acquired immunodeficiency syndrome or AIDS, metastatic cancer, and hematologic malignancy). The SAPS score is entered into a mathematical formula, which can be solved on a calculator and whose solution provides the predicted hospital mortality. Therefore no commercial computer software is necessary to perform this calculation. This simplicity plus its low cost have made SAPS a popular choice in some centers, particularly in Europe.
The Mortality Prediction Model (MPM).uses a mathematical formula whose solution provides a prediction of patient mortality. Typically, the MPM score is determined immediately upon ICU admission. The updated version of MPM (MPM24) uses a score twenty-four hours after ICU admission, utilizing five of the admission variables and eight additional physiologic variables. This provides two points of prognostic assessment within a 24-hour period. The MPM24 correlates with SAPS and APACHE, since all three are measured within twenty-four hours of ICU admission.
Many other predictive models have been developed for various purposes. By way of illustration and not as a limitation, a partial list of predictive models comprises SAPS II expanded and predicted mortality, SAPS II and predicted mortality, APACHE II and predicted mortality, SOFA (Sequential Organ Failure Assessment), MODS (Multiple Organ Dysfunction Score), ODIN (Organ Dysfunctions and/or INfection), MPM (Mortality Probability Model), MPM II LODS (Logistic Organ Dysfunction System), TRIOS (Three days Recalibrated ICU Outcome Score), EUROSCORE (cardiac surgery), ONTARIO (cardiac surgery), Parsonnet score (cardiac surgery), System 97 score (cardiac surgery), QMMI score (coronary surgery), Early mortality risk in redocoronary artery surgery, MPM for cancer patients, POSSUM (Physiologic and Operative Severity Score for the enUmeration of Mortality and Morbidity) (surgery, any), Portsmouth POSSUM (surgery, any), IRISS score: graft failure after lung transplantation, Glasgow Coma Score, ISS (Injury Severity Score), RTS (Revised Trauma Score), TRISS (Trauma Injury Severity Score), ASCOT (A Severity Characterization Of Trauma), 24 h—ICU Trauma Score, TISS (Therapeutic Intervention Scoring System), TISS-28 (simplified TISS), PRISM (Pediatric RISk of Mortality), P-MODS (Pediatric Multiple Organ Dysfunction Score), DORA (Dynamic Objective Risk Assesment), PELOD (Pediatric Logistic Organ Dysfunction), PIM II (Paediatric Index of Mortality II), PIM (Paediatric Index of Mortality), CRIB II (Clinical Risk Index for Babies), CRIB (Clinical Risk Index for Babies), SNAP (Score for Neonatal Acute Physiology), SNAP-PE (SNAP Perinatal Extension), SNAP II and SNAPPE II, MSSS (Meningococcal Septic Shock Score), GMSPS (Glasgow Meningococcal Septicaemia Prognostic Score), Rotterdam Score (meningococcal septic shock), Children's Coma Score (Raimondi), Paediatric Coma Scale (Simpson & Reilly), and Pediatric Trauma Score.
Predictive models are useful in establishing an initial treatment plan for a patient. Continuous assessment of patient specific rules can facilitate the implementation of the treatment plan. However, it would also desirable to monitor the progress of the patient using predictive models to evaluate the effectiveness of the treatment plan on a continuous basis and to revise the treatment plan and/or the patient specific rules accordingly.